krbdev
A workshop desk at night: eight blue-lit monitors and two laptops on articulated robot arms, a brass lamp, a cogwheel and a gauge clock.

Software engineer · Data scientist

Karlis Baisden

I build the software a business runs on, for the sales floor, the service shop, the plant floor, the front desk, and the people who run the business.

Microsoft Dynamics 365 and Power Platform, full-stack web apps, and the data and reporting behind them.

  • Led a food plant from paper to digital
  • 25+ years building software
  • Dynamics 365 and Power Platform, from first workshop to production

01 / What I build

Apps, automation, and reporting for the people who run the business.

I work on Microsoft Dynamics 365 and Power Platform, and on full-stack web apps: model-driven and canvas apps on Dataverse, Power Automate flows, reporting portals and the data behind them. The screens below are real demo modules, rebuilt from scratch at made-up companies with made-up data. Click one to use it full size.

D7Leadership Margin View

Units, sales, margin and trade-ins for any period across every branch, compared like for like, with alerts that act.

Live demo · made-up data

D2Unit Finder

Any unit at any branch, in stock or ever, as you type or from a pasted list.

Live demo · made-up data

D8Dashboard Launcher

One place inside the CRM where a rep or manager finds every role-based dashboard, grouped by job.

Live demo · made-up data

D1Service Floor

Which technicians are on a job right now, which are free, and who has not arrived.

Live demo · made-up data

D9Lead Routing Flow

The automation that hands each new lead to its branch team, starts the response clock and reports its own failures.

Live demo · made-up data

L1Conflicts Check

Every prior relationship with any party, including aliases and corporate family, as you type.

Live demo · made-up data

L2Matter Intake

One queue for every inquiry, owned by a team until claimed, on a business-hours clock.

Live demo · made-up data

L3Practice Financials

Fees worked, billed and collected for any period, compared like for like, with a bridge that ties to the cent.

Live demo · made-up data

02 / What shipped

What shipped, and what it looked like.

Some of the work from my current role at a multi-branch equipment dealer. Each one has a close recreation you can open and use. The companies and the data are made up.

01

Finding any unit

Before
Unit search didn't work for the people who needed it, and there were dozens of complaints a day.
What I built
One search over every unit at every branch, including all-time history (about half a million units), that filters as you type.
What changed
Replaced a report that took minutes with a search that answers in seconds.
Live recreation
02

Leadership reporting

What I built
About 20 report pages behind the company sign-in, each page shown only to the teams it's for.
Live recreation
03

The service floor

What I built
A live board of which technicians are on a job, free, or not in.
Live recreation
04

Lead routing

What I built
An automation that hands each new lead to its branch team, starts a response clock and reports its own failures.
Live recreation
05

Access and onboarding

What I built
New hires get their system access within about an hour of HR entering them, with nobody retyping IDs. Ad-hoc permissions were replaced by access by job role, set up the same way in every environment by one script.
No recreation yet

03 / Skills and tools

The tools I build with.

The sections above show what I built. This is what I built it with, and where each tool fits.

Languages

TypeScript for web apps and every demo on this page, Python and SQL for data work, PowerShell for admin and release scripts, Power Fx and DAX inside Power Platform.

  • TypeScript
  • JavaScript
  • Python
  • SQL
  • PowerShell
  • C# and .NET
  • Power Fx
  • DAX
  • HTML and CSS
  • Google Apps Script
  • PHP

Microsoft platform

Model-driven and canvas apps on Dataverse, the flows behind them, and the reporting on top.

  • Dynamics 365
  • Dataverse
  • Model-driven apps
  • Canvas apps
  • Power Automate
  • Power BI
  • Azure

Web apps

Full-stack web apps, reporting portals, and this site.

  • React
  • Next.js
  • Node.js
  • Vite
  • Tailwind CSS
  • Express
  • Zod
  • TanStack Query
  • Recharts
  • D3.js
  • Three.js

Databases

Where the data lives, from Dataverse tables to Postgres and BigQuery.

  • Dataverse
  • Azure SQL
  • PostgreSQL
  • Supabase
  • BigQuery
  • MySQL
  • Drizzle ORM

Data and reporting

Cleaning, modeling and charting data, from a notebook to the dashboard people read.

  • pandas
  • NumPy
  • SciPy
  • scikit-learn
  • Jupyter
  • dbt
  • Looker
  • Tableau
  • Streamlit
  • Plotly and Dash
  • matplotlib and seaborn

AI tools

I use AI coding tools every day. What they help write goes through the same review and release gate as everything else.

  • Claude Code
  • OpenAI Codex
  • Cursor
  • Claude API
  • OpenAI API
  • Grok (xAI)
  • Vercel AI SDK
  • MCP servers

Command line

Building, deploying and running things, mostly from a terminal.

  • Git
  • GitHub CLI
  • Power Platform CLI
  • Azure CLI
  • Vercel CLI
  • Google Cloud CLI
  • AWS CLI
  • Cloudflare Wrangler
  • Supabase CLI
  • pnpm
  • ffmpeg

Testing and hosting

Checking changes before they ship, and running them after.

  • Vitest
  • Playwright
  • Jest
  • CI/CD pipelines
  • Vercel
  • Cloudflare
  • AWS
  • Google Cloud
  • Docker

How this site is built

Next.js, React and TypeScript, styled with Tailwind CSS, checked by about 600 unit tests and 90 browser tests, and hosted on Vercel. The demos run on the same code, with made-up companies and data.

04 / How I work with people

I build with the people who will use the system, not for them.

Most of what makes a system work happens before and after the code. This is how I run a project.

01

Start at the desk where the work happens

I sit with the people who do the job today, watch how they do it, and write down the steps, the exceptions, and the workarounds. Requirements come from that, not from a form.

02

Show something working early

I run workshops with the stakeholders, then build a proof of concept and demo it while it is still rough. People react to a screen they can click far better than to a document.

03

Change it in short cycles

I work in short iterations with a visible backlog. What people say in the demo goes into the next cycle, and they see it again a week or two later.

04

Train, hand over, stay reachable

I write the guide, run the training, and stay close through the first weeks of real use. The system is done when the team runs it without me.

05 / How I run a platform

Releases, access, and AI, run the same careful way.

A platform is only as good as the care around it. These are the three rules I run every environment by.

Rule 01

Gated releases

Changes move from sandbox to production through a gate with an owner. Solutions travel in release packages through a pipeline under source control. Anything that cannot travel in a package is written down as a change entry, applied by a script, and checked afterwards.

SANDBOXGATEPRODUCTIONRELEASE GATE
Release gateSandbox
EntryWhatState
CE-0412Service · promised date columnapplied, checked
CE-0413Parts · backorder viewapplied, checked
CE-0414Flow · technician dispatchin review
Example change log
Rule 02

Access by job

Access follows the job, not the person. Each job gets a group with exactly the tables, apps, flows, and reports it needs, and one script sets the groups up the same way in every environment. Sandbox and production stay apart.

JobTablesAppsFlowsReports
Service writer
Parts counter
Technician
Finance
Example access by job
Rule 03

AI with a front door

Requests to use AI tools come in through one front door, and people are told when they will hear back. Code written with AI help goes through the same review and the same gate as any other change, and a person asks for and approves each one.

AskDecideBuildReviewGateProduction

Every change takes the same road

06 / Career

Where I have worked.

  1. Jan 2026 to nowCurrent
    Midwest Machinery Co.

    Software Engineer.

  2. 2025 to nowClient work
    Premier Studio

    Client builds.

  3. Sep 2025 to nowFounder
    Cogwheel.tech

    Founder and Solutions Engineer. Earlier, Founder and Solutions Consultant, Aug 2019 to Feb 2023.

  4. Feb 2023 to Sep 2025Food plant
    Seaboard Triumph Foods

    IT Technical Specialist. → Power Platform Developer. → Director of IT. Led the plant's move from paper to digital.

  5. 2000 to 2019Founder
    Coyote Web Solutions

    Founder and Web Developer.

07 / Education and credentials

Education and credentials.

A graduate degree in data science in progress, a business systems degree completed, and Microsoft's solution architect training on top of years of hands-on work.

Sc.M., Data Science

Brown University, online program

Sep 2025 to Dec 2026 · expected
B.S., Management Information Systems

Bellevue University, Dean's List

Completed Mar 2025
Network Administration and Security

Western Iowa Tech Community College, coursework

2019 to 2020

Credentials

Microsoft Power Platform Solution Architect (PL-600), completed training, 2025

PL-600T00 instructor-led course, certificate of achievement. Solution envisioning, architecture, security, release management, and integration for Dynamics 365 and Power Platform.

Course certificate
  • Microsoft
    Microsoft Certified: Power Platform Fundamentals2024
  • Google
    Google Project Management2025
  • University of Michigan
    Applied Machine Learning in Python2025
  • IBM
    Supervised Machine Learning: Classification2024
  • DataCamp
    SQL Associate2025
  • Illinois Tech
    Artificial IntelligenceIllinois Institute of Technology · 2025

08 / Contact

Let's talk.

If you are hiring for a senior Dynamics 365 or Power Platform role, or you have a system that needs building, I would like to hear about it. Email is the fastest way to reach me.

LinkedInGitHubEmail
Karlis Baisden
Tell me what you're building.
krbdev · Karlis Baisden